[Seaside] VW cgi interface

danil osipchuk danil at tsnet.ru
Fri Mar 26 10:19:40 CET 2004


Avi Bryant wrote:

> Yeah, I'd say so.  But there must be some other workaround, no?  Can you 
> maybe post a full test case (sql for setting up some tables and rows 
> with nulls, and then a roe-generated query that breaks) so that if there 
> are any SQL gurus out there they can suggest a fix?

RATestPostgresSemanticsWithNils does this. It adds additional column 
with nulls to every table and before testing removes it by projection to 
provide test consistency. One can set flag in 
RATestPostgresSemanticsWithNils>>addNilColumnsToTables: to fill 
additional column with 1 to check if trick is not too hackish (btw. 
testFindClassmates - suffers from it for some reason).

For example this statement fails (testUpdateSimpleSelect):
UPDATE students SET "name" = ('Fitzell') WHERE ("studentNumber", "name", 
"nilcolumn") IN (SELECT * FROM (SELECT "studentNumber" AS c1, "name" AS 
c2, "nilcolumn" AS c3 FROM students) AS t1 WHERE (c2='Julian'))

SELECT * FROM (SELECT "studentNumber" AS c1, "name" AS c2, "nilcolumn" 
AS c3 FROM students) AS t1 WHERE (c2='Julian')

gives: (2, 'Julian', NULL) and nothing can satisfy WHERE IN condition.

Right now I have no idea of general workaround not using semantics of 
existing table (primary keys, not null columns etc). So opinion of 
database gurus is very desirable.




More information about the Seaside mailing list